Bid invitation file review method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of machine learning, and discloses a bid invitation file examination method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining a to-be-examined target bid invitation file; determining at least one target rule item corresponding to the target bid inviting file; the target rule items are rule items in a preset rule item set, and at least part of the rule items in the rule item set are first-class rule items for reviewing the bid invitation file based on a large language model; positioning a target review object corresponding to the target rule item in the target bid invitation file; under the condition that the target rule item belongs to the first class of rule items, inputting the target review object into the large language model, and determining a review result of the target rule item according to the output of the large language model; and generating an examination result of the target bid invitation file. The bid invitation document can be quickly examined by using the large language model, and the efficiency is high; different rule items are used for examination, and all aspects of the bid invitation document can be comprehensively examined.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to a method, device, computer equipment and storage medium for reviewing bidding documents. Background Art

[0002] Bidding document review refers to a series of systematic and comprehensive inspections, evaluations and supervisions of bidding documents during bidding activities. Its purpose is to ensure that the preparation of bidding documents complies with the requirements of relevant laws and regulations, to ensure the standardization and legality of the bidding subject's behavior, to maintain a fair and competitive market environment, to improve the quality and efficiency of bidding work, and to promote the healthy and orderly development of the entire bidding market.

[0003] Currently, the review of bidding documents mainly relies heavily on manual work, which not only requires a lot of manpower costs, but is also easily affected by personal experience, leading to review errors or omissions, resulting in inaccurate review results. Summary of the invention

[0004] In view of this, the present invention provides a method, apparatus, computer equipment and storage medium for reviewing bidding documents, so as to enable rapid and accurate review of bidding documents.

[0005] In a first aspect, the present invention provides a method for reviewing a bidding document, comprising:

[0006] Obtain the target bidding documents for review;

[0007] Determine at least one target rule item corresponding to the target bidding document; the target rule item is a rule item in a preset rule item set, and at least some of the rule items in the rule item set are first-category rule items for reviewing the bidding document based on a large language model;

[0008] Locate the target review object corresponding to the target rule item in the target bidding document;

[0009] In the case where the target rule item belongs to the first category of rule items, inputting the target review object into the large language model, and determining the review result of the target rule item according to the output of the large language model;

[0010] The review results of the target bidding document are generated according to the review results of each of the target rule items.

[0011] In some optional implementations, after obtaining the target bidding document to be reviewed, the method further includes:

[0012] Parsing the bidding information to be reviewed from the target bidding document;

[0013] Performing structural analysis on the bidding information to determine chapter information corresponding to each chapter in the target bidding document;

[0014] Analyze the content of each chapter information respectively to determine the content of each paragraph in the corresponding chapter;

[0015] The locating the target review object corresponding to the target rule item in the target bidding document includes:

[0016] Determine the target paragraph content corresponding to the target rule item in each chapter;

[0017] The target paragraph content is used as the target review object corresponding to the target rule item.

[0018] In some optional embodiments, the method further comprises:

[0019] For the key chapters predefined in the target bidding document, extract key fields from the chapter information of the key chapters to determine key information corresponding to the key fields;

[0020] The step of taking the target paragraph content as a target review object corresponding to the target rule item includes:

[0021] The target paragraph content and the key information corresponding to the target rule item are used as the target review objects corresponding to the target rule item.

[0022] In some optional implementations, after determining the paragraph content of each paragraph in the corresponding chapter, the method further includes:

[0023] Determine summary information of the content of each of the paragraphs;

[0024] The determining of the target paragraph content corresponding to the target rule item in each chapter includes:

[0025] Determining whether the paragraph content matches the target rule item according to the summary information of the paragraph content;

[0026] In the case where the paragraph content matches the target rule item, the paragraph content is used as the target paragraph content corresponding to the target rule item.

[0027] In some optional implementations, the content parsing of each of the chapter information is performed to determine the content of each paragraph in the corresponding chapter, including:

[0028] Performing content analysis on each of the chapter information respectively to determine the paragraph content and paragraph position parameters of each paragraph in the corresponding chapter; the paragraph position parameters include the page number, position coordinates and width and height of the page to which the paragraph belongs;

[0029] The method further comprises:

[0030] In the case that there is a risk paragraph that has not passed the review in the target review object corresponding to the target rule item, the position of the risk paragraph in the target bidding document is determined according to the paragraph position parameter of the risk paragraph, and the risk paragraph is highlighted in the target bidding document.

[0031] In some optional embodiments, the method further comprises:

[0032] For the first type of rule items in the rule item set, historical review cases related to the first type of rule items in the local knowledge base are associated; the historical review cases are historical review objects and review results corresponding to the first type of rule items determined when reviewing previous historical bidding documents;

[0033] Associating laws and regulations related to the first category of rule items;

[0034] The step of inputting the target review object into the large language model and determining the review result of the target rule item according to the output of the large language model includes:

[0035] The target review object and historical review cases, laws and regulations associated with the target rule item are input into the large language model, and the review result of the target rule item is determined according to the output of the large language model.

[0036] In some optional implementations, some of the rule items in the rule item set are second-category rule items for reviewing the bidding documents based on other review algorithms other than the large language model;

[0037] The method further comprises:

[0038] In the case that the target rule item belongs to the second category of rule items, the target review object is reviewed according to the review algorithm corresponding to the target rule item to obtain the review result of the target rule item.

[0039] In some optional implementations, the determining at least one target rule item corresponding to the target bidding document includes:

[0040] Displaying multiple review rules belonging to the rule item set; the multiple review rules include: exclusive review rules, basic compliance review rules, consistency review rules, basic review rules and rationality review rules; each of the review rules includes at least one rule item;

[0041] In response to a selection instruction for selecting a review rule and / or a rule item in the review rule, the selected rule item is used as a target rule item.

[0042] In a second aspect, the present invention provides a device for reviewing a bidding document, comprising:

[0043] An acquisition module is used to acquire the target bidding documents to be reviewed;

[0044] A processing module is used to determine at least one target rule item corresponding to the target bidding document; the target rule item is a rule item in a preset rule item set, and at least some of the rule items in the rule item set are first-class rule items for reviewing bidding documents based on a large language model; locate a target review object corresponding to the target rule item in the target bidding document;

[0045] a review module, configured to input the target review object into the large language model when the target rule item belongs to the first category of rule items, and determine a review result of the target rule item according to an output of the large language model;

[0046] A generation module is used to generate the review results of the target bidding document according to the review results of each target rule item.

[0047] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the method for reviewing bidding documents of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0048] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for reviewing bidding documents of the first aspect or any corresponding embodiment thereof.

[0049] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the method for reviewing bidding documents of the first aspect or any corresponding embodiment thereof.

[0050] The present invention predefines a variety of rule items that need to be reviewed, and different rule items are used to review different contents of the bidding documents, and some of the rule items can review the bidding documents based on a large language model. By utilizing the powerful information processing capability of the large language model, a large amount of bidding documents can be quickly processed, thereby improving the review efficiency of the bidding documents; and the review result of the large language model is relatively more objective, making the review process more standardized and unified; by utilizing different rule items to review different contents of the bidding documents respectively, the influence between different rule items can be reduced, and while ensuring the accuracy of the review of each rule item, various aspects of the bidding documents can be relatively comprehensively covered, which is conducive to realizing a comprehensive review of the bidding documents. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related technologies, the drawings required for use in the specific embodiments or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 is a flow chart of a method for reviewing bidding documents according to an embodiment of the present invention;

[0053] Figure 2 is a flow chart of another method for reviewing bidding documents according to an embodiment of the present invention;

[0054] Figure 3 is an architectural diagram of a review system according to an embodiment of the present invention;

[0055] Figure 4 is a schematic diagram of reviewing a bidding document according to an embodiment of the present invention;

[0056] Figure 5 is a schematic diagram of a final review result according to an embodiment of the present invention;

[0057] Figure 6 is a schematic diagram of a positioning effect of an exclusive review result according to an embodiment of the present invention;

[0058] Figure 7 is a structural block diagram of a device for reviewing bidding documents according to an embodiment of the present invention;

[0059] Figure 8 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0061] After the bidding agent has completed the preparation of the bidding documents, it will use the traditional manual review method to conduct self-review and internal three-level review. This process will review whether there are format errors, typos, sensitive words, etc. in the bidding documents, and whether there are inconsistencies in the context, and then review whether the set bidding scope, qualification requirements, and bid evaluation methods are reasonable. After the self-review and internal three-level review, the tenderer will review whether the bidding documents cover the bidding requirements and scope completely, and whether there are any unfavorable clauses in the contract terms. After the tenderer's review, a supervisory review will be conducted. The supervisor will review the quality of the bidding documents to avoid complaints from bidders due to quality problems in the bidding documents. The bidding documents can only be finally released after passing all the above review processes. If any link fails to pass the review, it will be sent back to the superior review department or the original compiler, and then the review process will be repeated after modification. The whole process is very cumbersome and the review cycle is long.

[0062] This manual review of bidding documents has at least the following problems:

[0063] (1) Review the imbalance between manpower and workload.

[0064] Traditional review is highly dependent on manual operations. In actual work scenarios, the number of bidding documents that need to be reviewed is often extremely large, but the number of staff responsible for review is relatively scarce. This huge gap between manpower and workload makes it difficult to conduct a comprehensive and detailed review of all bidding documents under limited manpower conditions, which ultimately makes it difficult to achieve the goal of 100% full coverage of the review. For example, during the concentrated period of large-scale bidding projects, faced with a large number of bidding documents, even if the reviewers work overtime, they cannot guarantee that each document can be reviewed in a timely and complete manner. Some documents may be delayed or even missed due to insufficient manpower.

[0065] (2) The standards and processes for manual review are inconsistent.

[0066] Manual review mainly relies on the personal experience of auditors to judge the compliance and rationality of documents. Since personal experience is highly subjective and limited, different auditors may have different review standards and understanding of the same document, which leads to inconsistent overall review processes and standards. For example, an experienced auditor may be able to make quick judgments in familiar areas, but when faced with new business types or complex terms, it also takes a lot of time to study and analyze, and it is impossible to handle them as quickly as with standardized processes. Moreover, the accumulation of experience is a long process, and the lack of experience of new auditors during their growth will affect the overall review efficiency.

[0067] (3) Complex document review is prone to omissions.

[0068] When faced with bidding documents that are large or complex in content, the disadvantages of manual review become increasingly prominent. Such documents often contain numerous clauses, complex technical requirements, detailed business information, and a large number of attachments. During the long and intensive review process, auditors are prone to missing key information or misjudging the compliance of certain clauses due to factors such as fatigue and distraction. For example, a bidding document involving large-scale infrastructure construction may contain multiple complex parts such as engineering technical specifications, budget details, and contract terms. During manual review, some potential risk points or violations may be overlooked in the numerous details, thus laying hidden dangers for the subsequent implementation of the bidding project.

[0069] The method for reviewing bidding documents provided in an embodiment of the present invention predefines a variety of rule items that need to be reviewed. Different rule items are used to review different contents of bidding documents, and some of the rule items can review bidding documents based on a large language model. By using these rule items, format errors, logical contradictions, compliance issues, etc. in bidding documents can be automatically identified, thereby reducing omissions in manual review and improving review efficiency.

[0070] According to an embodiment of the present invention, an embodiment of a method for reviewing tender documents is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0071] In this embodiment, a method for reviewing bidding documents is provided, which can be applied to computers, mobile terminals, servers, etc. Figure 1 is a flow chart of a method for reviewing a bidding document according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps.

[0072] Step S101, obtaining the target bidding document to be reviewed.

[0073] In this embodiment, after the bidding documents are prepared, they can be reviewed. For the convenience of distinguishing and describing, these bidding documents that need to be reviewed are referred to as target bidding documents.

[0074] When a user needs to review a bidding document, he or she can actively upload and provide the bidding document, thereby obtaining the corresponding target bidding document. For example, a review system capable of implementing the method can be provided, and the user can upload the bidding document to be reviewed (i.e., the target bidding document) to the system; wherein, multiple formats of bidding documents can be supported, for example, the formats of the bidding documents can be pdf, doc, docx, etc.

[0075] Step S102, determining at least one target rule item corresponding to the target bidding document; the target rule item is a rule item in a preset rule item set, and at least some of the rule items in the rule item set are first-category rule items for reviewing the bidding document based on a large language model.

[0076] In this embodiment, multiple rule items are pre-set, each rule item is used to review the corresponding content in the bidding document, and these rule items can form a corresponding set, that is, a rule item set. For example, a rule item is used to check for typos in the bidding document, and another rule item is used to review whether the format of the bidding document is correct, etc.

[0077] Among them, some or all of the rule items in the rule item set are reviewed based on the large language model (LLM), that is, the content related to the rule item in the bidding document is reviewed based on the preset large language model; for the convenience of description, the rule items reviewed using the large language model are called first-class rule items. For example, if a rule item is to review the rationality of the bidding method based on the large language model, then this rule item is a first-class rule item.

[0078] For the target bidding document to be reviewed, it is possible to determine which contents of the target bidding document need to be reviewed, and then use the relevant rule items as the rule items that need to be used currently, namely, the target rule items.

[0079] Among them, which rule items are target rule items can be automatically determined, for example, each rule item in the rule item set is used as a target rule item, or the relevant target rule items are determined based on the user's current review needs. Alternatively, the user can actively select the rule items that need to be reviewed, and these rule items are the target rule items.

[0080] Step S103, locating the target review object corresponding to the target rule item in the target bidding document.

[0081] In this embodiment, for each rule item, it corresponds to a corresponding review object, which is the content in the bidding document that needs to be reviewed based on the rule item. Accordingly, when reviewing the current target bidding document based on the target rule item, it is necessary to determine the review object corresponding to the target rule item in the target bidding document, that is, the target review object.

[0082] For example, if a target rule item is used to review the qualifications of bidders, the content related to the qualifications of bidders in the target bidding document can be used as the target review object. Alternatively, if a target rule item is used to review typos, all the content in the target bidding document can be used as the target review object.

[0083] Step S104: when the target rule item belongs to the first category of rule items, the target review object is input into the large language model, and the review result of the target rule item is determined according to the output of the large language model.

[0084] In this embodiment, if a target rule item belongs to the first category of rule items, that is, the target rule item needs to be reviewed using a large language model, then the target review object of the target rule item is input into the large language model, and the target review object is reviewed using the powerful information processing capabilities of the large language model. The review results can be obtained quickly and accurately, effectively freeing up manpower.

[0085] Among them, a large language model suitable for the bidding field is pre-trained, and by calling the large language model, the use of the large language model can be realized. If multiple target rule items all belong to the first category of rule items, the large language model can be called simultaneously for each target rule item, thereby realizing the simultaneous review of multiple target rule items.

[0086] Specifically, for each first-category rule item, a corresponding prompt word can be set in advance; if the target rule item belongs to the first-category rule item, the target review object and the prompt word corresponding to the target rule item can be input into the large language model together, so that the large language model can more accurately complete the review work of the target review object, thereby obtaining a more accurate review result.

[0087] In addition, optionally, for some contents in the bidding documents, if the review method is relatively simple, the large language model may not be used, but the traditional algorithm may be used as the review algorithm. Specifically, some rule items in the rule item set are second-category rule items that review the bidding documents based on review algorithms other than the large language model; that is, some rule items are first-category rule items that use the large language model, and another part of the rule items are second-category rule items that use the traditional algorithm.

[0088] For example, if a rule item is reviewed based on a traditional matching algorithm (such as regular expressions, etc.) or based on a preset text vectorization model, these rule items do not need to call a large language model and belong to the second category of rule items.

[0089] Correspondingly, the method also includes: when the target rule item belongs to the second category of rule items, reviewing the target review object according to the review algorithm corresponding to the target rule item to obtain the review result of the target rule item.

[0090] In this embodiment, for certain target rule items, if they belong to the second category of rule items, that is, these target rule items are rule items using traditional algorithms, then the target review object is reviewed based on the corresponding review algorithm to obtain the corresponding review result.

[0091] Step S105, generating the review result of the target bidding document according to the review result of each target rule item.

[0092] In this embodiment, the review results of each target rule item are summarized to obtain the overall review result of the target bidding document; for example, a total review report can be generated in combination with the review results of each target rule item, so that the user can easily identify the unreasonable parts in the target bidding document by viewing the review report. Each review result can include a detailed review conclusion of the target review object (such as whether it is abnormal, etc.), and for the target review object with errors, the reasons and modification suggestions can also be further given.

[0093] The review method for the bidding documents provided in this embodiment predefines a variety of rule items that need to be reviewed. Different rule items are used to review different contents of the bidding documents, and some of the rule items can review the bidding documents based on a large language model. By utilizing the powerful information processing capabilities of the large language model, a large amount of bidding documents can be quickly processed, thereby improving the review efficiency of the bidding documents. Moreover, the review results of the large language model are relatively more objective, making the review process more standardized and unified. By utilizing different rule items to review different contents of the bidding documents respectively, the influence between different rule items can be reduced. While ensuring the accuracy of the review of each rule item, all aspects of the bidding documents can be relatively comprehensively covered, which is conducive to achieving a comprehensive review of the bidding documents.

[0094] In this embodiment, a method for reviewing bidding documents is provided, which can be applied to computers, mobile terminals, servers, etc. Figure 2 is a flow chart of a method for reviewing a bidding document according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps.

[0095] Step S201, obtaining the target bidding document to be reviewed.

[0096] For details, please see Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0097] The target bidding document needs to be parsed first to extract the content of the target bidding document, so as to determine the review object corresponding to each rule item. The process of parsing the target bidding document may include the following steps S202 to S204.

[0098] Step S202, parsing the bidding information to be reviewed from the target bidding document.

[0099] In this embodiment, by parsing the target bidding document, the content that needs to be reviewed, namely, the bidding information, can be extracted. The bidding information is mainly information in text form.

[0100] Parsing components for processing bidding documents in different formats can be preset. For example, depending on the specific format differences of the bidding documents, the parsing components adapted to them are enabled. For example, if the bidding document is in PDF format, a special PDF parsing component is called; and for bidding documents in doc or docx format, a word file parsing component is enabled. After being processed by the corresponding parsing components, the bidding information in the bidding documents can be extracted so that the subsequent review process can be carried out smoothly and efficiently.

[0101] Step S203, performing structural analysis on the bidding information to determine the chapter information corresponding to each chapter in the target bidding document.

[0102] In this embodiment, since the bidding documents generally have a relatively fixed format, the paragraph structure of the bidding documents is relatively uniform. By performing structural analysis on the bidding information extracted from the target bidding documents, the chapters in the target bidding documents can be determined and divided, and then the information corresponding to each chapter, i.e., the chapter information, can be determined.

[0103] Specifically, a unified processing flow is used to parse the detailed contents of the bidding documents. In the initial stage of the parsing process, the paragraph structure of the bidding documents is analyzed. With the help of recognition technology and splitting methods, specific chapter structure categories can be separated. These chapters can generally include: bidding notice, bidder instructions, bid evaluation method, contract terms and format, bill of quantities, drawings, technical standards and requirements, and bidding document format.

[0104] Through structured analysis, the structure of the bidding documents can be effectively deconstructed and detailed, laying the foundation for the subsequent in-depth processing and information extraction of each paragraph structure.

[0105] Step S204, respectively analyze the content of each chapter information to determine the paragraph content of each paragraph in the corresponding chapter.

[0106] In this embodiment, each chapter in the bidding document generally includes one or more layers of titles, and each title further includes one or more paragraphs; by parsing the content of the corresponding chapter, the paragraphs in the corresponding chapter can be divided, and the content contained in each paragraph, i.e., the paragraph content, can be determined. It can be understood that the paragraph content is part of the content in the chapter information.

[0107] Step S205, determining at least one target rule item corresponding to the target bidding document; the target rule item is a rule item in a preset rule item set, and at least some of the rule items in the rule item set are first-category rule items for reviewing the bidding document based on a large language model.

[0108] For details, please see Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0109] Optionally, the user may actively select the desired target rule item. Specifically, the above step S205 "determine at least one target rule item corresponding to the target bidding document" may include steps A1 to A2.

[0110] Step A1, display multiple review rules belonging to the rule item set; the multiple review rules include: exclusive review rules, basic compliance review rules, consistency review rules, basic review rules and rationality review rules; each review rule includes at least one rule item.

[0111] Step A2, in response to an instruction to select a review rule and / or a rule item in the review rule, the selected rule item is used as a target rule item.

[0112] In this embodiment, in order to reasonably set rule items and facilitate user selection and use, a variety of review rules for bid documents are summarized through in-depth analysis of the bidding business; and each review rule includes at least one rule item. Generally, each review rule includes multiple rule items, and these rule items form a rule item set.

[0113] In this embodiment, at least five types of review rules are divided, including: exclusive review rules, basic compliance review rules, consistency review rules, basic review rules and rationality review rules.

[0114] (1) Exclusive review: Use AI (artificial intelligence) technologies such as large language models to conduct a targeted review of the content of the bidding documents, including regional requirements, localization requirements, ownership requirements, product origin requirements, specific awards, performance, etc., to ensure the fairness and objectivity of the bidding process.

[0115] (2) Basic compliance review: Based on the requirements of relevant laws and regulations (such as the Implementing Regulations of the Tendering and Bidding Law, etc.), the full text of the tender documents is scanned for compliance, including time requirements, access conditions, bid setting rights, etc., to ensure that the content of the tender documents complies with the requirements of laws and regulations.

[0116] (3) Consistency review: Scan the full text of the bidding documents based on the text parsing capabilities of the large language model to ensure the consistency of the context of the same matter, the consistency of the bidding documents with the sample text, etc.

[0117] (4) Basic review: Based on the semantic understanding capability of the large language model and the construction of a sensitive word library, the full text of the bidding documents is scanned to ensure that there are no obvious typos or sensitive words.

[0118] (5) Reasonableness review: Based on policy requirements, the full text of the bidding documents is scanned for reasonableness, including the requirement to provide original documents without transition, the number of evaluation experts, and the requirements for consortiums in joint bidding.

[0119] When the user selects the required rule item, the above five review rules can be displayed, and the drop-down menu of each review rule includes the corresponding rule item. The user can trigger the selection instruction to select one or some review rules, at which time all the rule items under the selected review rule can be used as the target rule item; or the user can also select the currently required rule item from the various rule items under a certain review rule, and the rule item selected by the user is the target rule item.

[0120] Alternatively, the user may select all review rules, that is, all rule items are used as target rule items.

[0121] For example, after the user uploads the bidding documents to be reviewed (i.e., the target bidding documents), he or she can flexibly configure the specific rule items for this review. He or she can select all review rules to conduct a comprehensive and detailed review of the bidding documents and identify all possible problems. He or she can also select a certain review rule or part of the rule items to conduct a focused review of the bidding documents and identify possible problems with the selected rule items.

[0122] in, Figure 3 An architectural diagram of the review system for implementing the method provided by this embodiment is shown; Figure 3 As shown, by setting traditional review algorithms such as a large language model (referred to as the big model) and a text vectorization model, after the bidding documents are parsed, the bidding documents can be reviewed according to corresponding rules and finally an inspection result (i.e., a review result) can be generated.

[0123] Step S206, locating the target review object corresponding to the target rule item in the target bidding document.

[0124] Specifically, the above step S206 "locating the target review object corresponding to the target rule item in the target bidding document" includes steps S2061 to S2062.

[0125] Step S2061, determining the target paragraph content corresponding to the target rule item in each chapter.

[0126] In this embodiment, each paragraph of each chapter may correspond to one or more rule items, wherein the paragraph content can be determined based on the corresponding paragraph content to which rule item or items the paragraph content corresponds; the paragraph content corresponding to the target rule item is referred to as the target paragraph content.

[0127] Optionally, after the above step S204 of "determining the paragraph content of each paragraph in the corresponding chapter", the method further includes step B1.

[0128] Step B1, determining summary information of each paragraph content.

[0129] Furthermore, the above step S2061 of "determining the target paragraph content corresponding to the target rule item in each chapter" may include steps C1 to C2.

[0130] Step C1, judging whether the paragraph content matches the target rule item according to the summary information of the paragraph content.

[0131] Step C2: when the paragraph content matches the target rule item, the paragraph content is used as the target paragraph content corresponding to the target rule item.

[0132] In this embodiment, for each paragraph content, summary information can be extracted. When determining which rule items the paragraph content corresponds to, the determination is made based on the summary information of the paragraph content.

[0133] Specifically, each rule item predefines the relevant review business and preconfigures the review process corresponding to the rule item, for example, whether the rule item is reviewed based on a large language model or a traditional algorithm. In addition, for the paragraph content of a paragraph, its similarity with the review business of the target rule item can be determined. If the similarity between the two exceeds a preset threshold, the paragraph content can be considered to match the target rule item, and the paragraph content can be used as the target paragraph content corresponding to the target rule item.

[0134] For example, the hierarchical results between titles and paragraphs in a chapter can be analyzed, and the subtitle of the current paragraph can be used as the summary information of the current paragraph; and the title of the previous level of the current chapter can also be used as the summary information of the current level title and the current paragraph. By executing the above processing flow, it is possible to construct a mapping association between the titles of each level in the entire chapter and the corresponding paragraphs, and then generate complete and relevant summary information, providing accurate and structured basic data support for subsequent data processing, information retrieval, content summarization and other operations.

[0135] Step S2062, taking the target paragraph content as the target review object corresponding to the target rule item.

[0136] In this embodiment, since the target paragraph content is related to the target rule item, that is, the target paragraph content contains content that needs to be reviewed by the target rule item, it is necessary to use the target paragraph content as the target review object corresponding to the target rule item so that the target paragraph content can be subsequently reviewed based on the review algorithm of the target rule item.

[0137] In addition, optionally, after the above step S203 of "performing structural analysis on the bidding information to determine the chapter information corresponding to each chapter in the target bidding document", the method further includes the following step B2.

[0138] Step B2, for the key chapters predefined in the target bidding document, extract key fields from the chapter information of the key chapters to determine key information corresponding to the key fields.

[0139] Furthermore, the above-mentioned step S2062 "using the target paragraph content as the target review object corresponding to the target rule item" may include: using the target paragraph content and the key information corresponding to the target rule item as the target review object corresponding to the target rule item.

[0140] In this embodiment, some chapters of the bidding document are used as key chapters. The key chapters are predefined, that is, which chapters are key chapters can be set in advance; for example, chapters such as "Bidding Notice", "Instructions for Bidders", and "Bidding Evaluation Methods" can be used as key chapters. Among them, the key chapters can be automatically predefined by the system, or the user can actively select which chapters to use as key chapters, that is, the user manually predefines which chapters are key chapters, which is not limited in this embodiment.

[0141] For the chapter information of these key chapters, we can use technical means such as large language models to accurately extract the key information in the bidding process.

[0142] Specifically, the key fields to be extracted in each key chapter can be pre-set, and information can be extracted based on the key fields to determine the corresponding key information. For example, the key fields of the bidding announcement chapter can include multiple key fields such as project name, bid section name, funding source, and investment ratio, so as to extract the content of each key field to form key information.

[0143] In the subsequent review process, in addition to the target paragraph content, the corresponding key information will also be reviewed, so that corresponding review work can be carried out on these extracted key information to ensure the compliance, accuracy and rigor of the entire bidding process.

[0144] Step S207: When the target rule item belongs to the first category of rule items, the target review object is input into the large language model, and the review result of the target rule item is determined according to the output of the large language model.

[0145] For details, please see Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0146] In some optional implementations, for each preset first-category rule item, background knowledge that needs to be input into the large language model may be preset. Specifically, the method may further include the following steps D1 and D2.

[0147] Step D1, for the first category of rule items in the rule item set, associating the historical review cases related to the first category of rule items in the local knowledge base; the historical review cases are the historical review objects and review results corresponding to the first category of rule items determined when reviewing the previous historical bidding documents;

[0148] Step D2, associating laws and regulations related to the first category of rule items.

[0149] Moreover, when the target rule item belongs to the first category of rule items, the above-mentioned step S207 "inputting the target review object into the large language model, and determining the review result of the target rule item according to the output of the large language model" may specifically include: inputting the target review object and historical review cases, laws and regulations associated with the target rule item into the large language model, and determining the review result of the target rule item according to the output of the large language model.

[0150] In this embodiment, for the first category of rule items in the rule item set, in order to obtain more accurate review results using the large language model, previous review cases related to the first category of rule items, i.e., historical review cases, are determined from the local knowledge base for storing review cases. Specifically, according to the way that the first category of rule items are reviewed using the large language model, the historical review objects related to the first category of rule items in the historical bidding documents are reviewed, and corresponding review results can be obtained, thereby forming historical review cases of the historical bidding documents and storing them in the local knowledge base.

[0151] In addition, the historical review cases related to the first category of rule items in the local knowledge base are pre-associated with the first category of rule items, so that when other bidding documents (target bidding documents) are subsequently reviewed based on the first category of rule items, the associated historical review cases can be directly used as background knowledge of the first category of rule items and input into the large language model, so that the large language model can more accurately review the target review objects in the target bidding documents.

[0152] Similarly, the first category of rule items are pre-associated with the relevant laws and regulations required by the rules, thereby preparing for subsequent review.

[0153] For example, some exclusive review rules are set based on the Fair Competition Review Regulations, so the relevant laws and regulations in the Fair Competition Review Regulations can be linked to these exclusive review rules and serve as corresponding background knowledge.

[0154] It can be understood that if the target rule item is a first-category rule item, then when conducting a review based on the target rule item, in addition to determining the target review object, historical review cases and laws and regulations associated with the target rule item can also be determined, and the historical review cases and laws and regulations can be used as background knowledge and input into the large language model together with the target review object, so that the large language model can refer to this background knowledge and conduct a more accurate review of the target review object.

[0155] Step S208: Generate the review result of the target bidding document according to the review result of each target rule item.

[0156] For details, please see Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0157] Figure 4 A schematic diagram showing the review of the bidding documents is shown in Figure 1. Figure 4As shown, for the bidding documents that need to be reviewed, they are first parsed to extract the contents of each chapter and the paragraphs under the chapters, and the key information corresponding to the key fields in the key chapters is extracted; and the user can select the required rule items, i.e., the target rule items, from the preset review rules.

[0158] For each target rule item, the corresponding review process can be matched according to the business definition of the review rule to determine the required review algorithm; for example, if the user selects the exclusive rule, the exclusive rule processing process based on the large model will be matched, and if the user selects the consistency rule, the consistency rule processing process based on the traditional algorithm will be matched. In addition, according to the business definition of the review rule and the algorithm requirements of the rule processing process, the key information corresponding to the specific paragraphs and specific key fields after the bidding document is parsed is located, and combined with the related historical review cases and relevant laws and regulations, preparations are made for subsequent reviews.

[0159] When conducting a review based on each target rule item, a specific review algorithm is matched against the specific review rule, and then the review object, related historical review case information, related legal and regulatory information, etc. are input into the review algorithm. The review algorithm finally outputs the review result after comprehensive processing.

[0160] After completing the review of all review rules, summarize the review results of each review rule item and output the review report. The review report contains the overall review conclusion and supports page browsing and export. For example, this time a total of 4 categories and 10 items were reviewed (one category for each review rule), of which there were 2 risk issues. At the same time, the report can also include detailed review conclusions, that is, the review conclusions corresponding to each rule, whether there is a risk, and if there is a risk, the risk cause and modification suggestions are prompted.

[0161] Figure 5 A schematic diagram showing the final review results. Figure 5 As shown, a total of 11 items in 5 categories were reviewed, including 10 risk issues, and some rule items corresponding to various review rules were schematically shown.

[0162] Optionally, the above step S204 "perform content analysis on each chapter information separately to determine the paragraph content of each paragraph in the corresponding chapter" may specifically include: perform content analysis on each chapter information separately to determine the paragraph content and paragraph position parameters of each paragraph in the corresponding chapter; the paragraph position parameters include the page number, position coordinates and width and height of the page to which the paragraph belongs.

[0163] Furthermore, the method further comprises step E1.

[0164] Step E1, when there is a risk paragraph that has not passed the review in the target review object corresponding to the target rule item, determine the position of the risk paragraph in the target bidding document according to the paragraph position parameter of the risk paragraph, and highlight the risk paragraph in the target bidding document.

[0165] In this embodiment, when parsing each paragraph, in addition to determining the paragraph content, a parameter that can indicate the location of the paragraph, namely, a paragraph location parameter, is also determined. The paragraph location parameter includes the page number, location coordinates, and width and height of the page to which the paragraph belongs. In the subsequent review process, if it is found that there are risky paragraphs that have not passed the review in the target review object being reviewed, the risky paragraphs can be located according to the paragraph location parameters and the risky paragraphs can be highlighted.

[0166] Specifically, for a risk paragraph, it is possible to determine which page (i.e., the page to which it belongs) in the target bidding document the risk paragraph is located based on the page number to which it belongs, and determine the position of the risk paragraph in the page to which it belongs based on the position coordinates, and determine a bounding box that can select the risk paragraph based on the width and height of the page to which it belongs, and highlight the risk paragraph based on the bounding box or highlight it in other colors.

[0167] For example, for Figure 5 The review results shown are Figure 6 A schematic diagram showing the positioning effect of the exclusive examination results is shown. Figure 6 As shown, after exclusive review, it was determined that the qualification settings in the bidding documents were unreasonable, and the risk reasons and corresponding modification suggestions were given for user reference.

[0168] The embodiment of the present invention uses a large language model to perform in-depth semantic analysis and key information extraction on the bidding documents, and extracts information directly related to the review rules from the complex content text of the bidding documents. In addition, for multiple rule items, a multi-threaded high-concurrency approach can be used to review different review rules at the same time. For specific review rules, the corresponding review algorithm is intelligently matched to identify the specific paragraphs in the bidding documents that need to be reviewed, and at the same time match the corresponding historical cases and relevant laws and regulations. Through comprehensive processing of the above matters, the intelligent review results are obtained and finally a review report is generated.

[0169] By utilizing the powerful information processing capabilities of the big model, it is possible to quickly process massive amounts of bidding documents, and simultaneously process different review rules for different bidding documents in a multi-threaded and high-concurrency manner, effectively making up for the lack of manpower in manual review when faced with a huge workload, improving the efficiency of the review, and ensuring that there will be no delayed or missed review of documents.

[0170] The big model establishes relatively objective and standardized audit rules and models by studying and analyzing a large amount of compliant and non-compliant bidding document data. It can eliminate the problem of inconsistent audit standards caused by differences in personal experience of different auditors, making the audit process more standardized and unified, which is conducive to building unified audit standards.

[0171] In addition, the big model has the ability of deep semantic understanding and multi-dimensional information integration. For bidding documents with large volume and complex content, it can comprehensively and accurately analyze the various parts such as engineering technical specifications, budget details, contract terms and their interrelationships, and effectively explore potential risk points and violations, thereby greatly improving the accuracy of complex document review, reducing the hidden dangers caused by omissions in the review to the subsequent implementation of the bidding project, and ensuring the smooth progress and quality safety of the bidding project.

[0172] In this embodiment, a device for reviewing bidding documents is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0173] This embodiment provides a device for reviewing bidding documents, such as Figure 7 As shown, including:

[0174] An acquisition module 701 is used to acquire a target bidding document to be reviewed;

[0175] The processing module 702 is used to determine at least one target rule item corresponding to the target bidding document; the target rule item is a rule item in a preset rule item set, and at least some of the rule items in the rule item set are first-class rule items for reviewing the bidding document based on a large language model; locate the target review object corresponding to the target rule item in the target bidding document;

[0176] A review module 703, configured to input the target review object into the large language model, and determine a review result of the target rule item according to an output of the large language model, when the target rule item belongs to the first category of rule items;

[0177] The generating module 704 is used to generate the review result of the target bidding document according to the review result of each target rule item.

[0178] In some optional implementations, after obtaining the target bidding document to be reviewed, the processing module 702 is further configured to:

[0179] Parsing the bidding information to be reviewed from the target bidding document;

[0180] Performing structural analysis on the bidding information to determine chapter information corresponding to each chapter in the target bidding document;

[0181] Analyze the content of each chapter information respectively to determine the content of each paragraph in the corresponding chapter;

[0182] The locating the target review object corresponding to the target rule item in the target bidding document includes:

[0183] Determine the target paragraph content corresponding to the target rule item in each chapter;

[0184] The target paragraph content is used as the target review object corresponding to the target rule item.

[0185] In some optional implementations, the processing module 702 is further configured to:

[0186] For the key chapters predefined in the target bidding document, extract key fields from the chapter information of the key chapters to determine key information corresponding to the key fields;

[0187] The processing module 702 uses the target paragraph content as a target review object corresponding to the target rule item, including:

[0188] The target paragraph content and the key information corresponding to the target rule item are used as the target review objects corresponding to the target rule item.

[0189] In some optional implementations, after determining the paragraph content of each paragraph in the corresponding chapter, the processing module 702 is further configured to:

[0190] Determine summary information of the content of each of the paragraphs;

[0191] The processing module 702 determines the target paragraph content corresponding to the target rule item in each chapter, including:

[0192] Determining whether the paragraph content matches the target rule item according to the summary information of the paragraph content;

[0193] In the case where the paragraph content matches the target rule item, the paragraph content is used as the target paragraph content corresponding to the target rule item.

[0194] In some optional implementations, the processing module 702 performs content analysis on each of the chapter information to determine the content of each paragraph in the corresponding chapter, including:

[0195] Performing content analysis on each of the chapter information respectively to determine the paragraph content and paragraph position parameters of each paragraph in the corresponding chapter; the paragraph position parameters include the page number, position coordinates and width and height of the page to which the paragraph belongs;

[0196] The generating module 704 is further used for:

[0197] In the case that there is a risk paragraph that has not passed the review in the target review object corresponding to the target rule item, the position of the risk paragraph in the target bidding document is determined according to the paragraph position parameter of the risk paragraph, and the risk paragraph is highlighted in the target bidding document.

[0198] In some optional implementations, the processing module 702 is further configured to:

[0199] For the first type of rule items in the rule item set, historical review cases related to the first type of rule items in the local knowledge base are associated; the historical review cases are historical review objects and review results corresponding to the first type of rule items determined when reviewing previous historical bidding documents;

[0200] Associating laws and regulations related to the first category of rule items;

[0201] The review module 703 inputs the target review object into the large language model, and determines the review result of the target rule item according to the output of the large language model, including:

[0202] The target review object and historical review cases, laws and regulations associated with the target rule item are input into the large language model, and the review result of the target rule item is determined according to the output of the large language model.

[0203] In some optional implementations, some of the rule items in the rule item set are second-category rule items for reviewing the bidding documents based on other review algorithms other than the large language model;

[0204] The review module 703 is also used for:

[0205] In the case that the target rule item belongs to the second category of rule items, the target review object is reviewed according to the review algorithm corresponding to the target rule item to obtain the review result of the target rule item.

[0206] In some optional implementations, the processing module 702 determines at least one target rule item corresponding to the target bidding document, including:

[0207] Displaying multiple review rules belonging to the rule item set; the multiple review rules include: exclusive review rules, basic compliance review rules, consistency review rules, basic review rules and rationality review rules; each of the review rules includes at least one rule item;

[0208] In response to a selection instruction for selecting a review rule and / or a rule item in the review rule, the selected rule item is used as a target rule item.

[0209] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0210] The review device for the bidding documents in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, including a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0211] The embodiment of the present invention also provides a computer device having the above Figure 8 The device for reviewing the tender documents is shown.

[0212] See also Figure 8 , Figure 8 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 8 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 A processor 10 is taken as an example.

[0213] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0214] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0215] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0216] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0217] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0218] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0219] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0220] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations should all be included in the protection scope of the present invention.

Claims

1. A method for reviewing bidding documents, characterized in that: The method comprises: Obtain the target bidding documents for review; Determine at least one target rule item corresponding to the target bidding document; the target rule item is a rule item in a preset rule item set, and at least some of the rule items in the rule item set are first-category rule items for reviewing the bidding document based on a large language model; Locate the target review object corresponding to the target rule item in the target bidding document; In the case where the target rule item belongs to the first category of rule items, inputting the target review object into the large language model, and determining the review result of the target rule item according to the output of the large language model; The review results of the target bidding document are generated according to the review results of each of the target rule items.

2. The method according to claim 1, characterized in that After obtaining the target bidding document to be reviewed, the method further includes: Parsing the bidding information to be reviewed from the target bidding document; Performing structural analysis on the bidding information to determine chapter information corresponding to each chapter in the target bidding document; Analyze the content of each chapter information respectively to determine the content of each paragraph in the corresponding chapter; The locating the target review object corresponding to the target rule item in the target bidding document includes: Determine the target paragraph content corresponding to the target rule item in each chapter; The target paragraph content is used as the target review object corresponding to the target rule item.

3. The method according to claim 2, characterized in that Also includes: For the key chapters predefined in the target bidding document, extract key fields from the chapter information of the key chapters to determine key information corresponding to the key fields; The step of taking the target paragraph content as a target review object corresponding to the target rule item includes: The target paragraph content and the key information corresponding to the target rule item are used as the target review objects corresponding to the target rule item.

4. The method according to claim 2, characterized in that: After determining the paragraph content of each paragraph in the corresponding chapter, the method further includes: Determine summary information of the content of each of the paragraphs; The determining of the target paragraph content corresponding to the target rule item in each chapter includes: Determining whether the paragraph content matches the target rule item according to the summary information of the paragraph content; In the case where the paragraph content matches the target rule item, the paragraph content is used as the target paragraph content corresponding to the target rule item.

5. The method according to claim 2, characterized in that: The content analysis of each of the chapter information is performed to determine the content of each paragraph in the corresponding chapter, including: Performing content analysis on each of the chapter information respectively to determine the paragraph content and paragraph position parameters of each paragraph in the corresponding chapter; the paragraph position parameters include the page number, position coordinates and width and height of the page to which the paragraph belongs; The method further comprises: In the case that there is a risk paragraph that has not passed the review in the target review object corresponding to the target rule item, the position of the risk paragraph in the target bidding document is determined according to the paragraph position parameter of the risk paragraph, and the risk paragraph is highlighted in the target bidding document.

6. The method according to claim 1, characterized in that The method further comprises: For the first type of rule items in the rule item set, historical review cases related to the first type of rule items in the local knowledge base are associated; the historical review cases are historical review objects and review results corresponding to the first type of rule items determined when reviewing previous historical bidding documents; Associating laws and regulations related to the first category of rule items; The step of inputting the target review object into the large language model and determining the review result of the target rule item according to the output of the large language model includes: The target review object and historical review cases, laws and regulations associated with the target rule item are input into the large language model, and the review result of the target rule item is determined according to the output of the large language model.

7. The method according to claim 1, characterized in that Some of the rule items in the rule item set are second-category rule items for reviewing the bidding documents based on other review algorithms other than the large language model; The method further comprises: In the case that the target rule item belongs to the second category of rule items, the target review object is reviewed according to the review algorithm corresponding to the target rule item to obtain the review result of the target rule item.

8. The method according to claim 1, characterized in that The determining of at least one target rule item corresponding to the target bidding document includes: Displaying multiple review rules belonging to the rule item set; the multiple review rules include: exclusive review rules, basic compliance review rules, consistency review rules, basic review rules and rationality review rules; each of the review rules includes at least one rule item; In response to a selection instruction for selecting a review rule and / or a rule item in the review rule, the selected rule item is used as a target rule item.

9. A device for reviewing bidding documents, characterized in that: The device comprises: An acquisition module is used to acquire the target bidding documents to be reviewed; A processing module is used to determine at least one target rule item corresponding to the target bidding document; the target rule item is a rule item in a preset rule item set, and at least some of the rule items in the rule item set are first-class rule items for reviewing bidding documents based on a large language model; locate a target review object corresponding to the target rule item in the target bidding document; a review module, configured to input the target review object into the large language model when the target rule item belongs to the first category of rule items, and determine a review result of the target rule item according to an output of the large language model; A generation module is used to generate the review results of the target bidding document according to the review results of each target rule item.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for reviewing tender documents according to any one of claims 1 to 8 by executing the computer instructions.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for reviewing bidding documents according to any one of claims 1 to 8.

Images